The advent of deep learning in computer vision has enabled the detection and segmentation of plant phenotypic traits with increased accuracy and efficiency. The challenges involved in segmenting the individual instances of plant traits using traditional image processing can be addressed with the help of supervised instance segmentation techniques such as Mask R-CNN. This thesis presents two instance segmentation applications to extract traits from 2-D images of corn cob and cotton plant. The former develops a deep learning-based image processing pipeline that aims to estimate the consumption of corn by identifying corn and its bare ear, which will aid in testing the wild animals’ preference for genetically modified corn. The estimation results of these models were included and compared with manually labeled test data with R2 = 0.99, which showed that the use of Mask R-CNN model provides highly accurate results, thus, allowing it to be used further on all collected data. The later study uses Mask R-CNN for extracting three of the most crucial cotton plant phenotypic traits: main stalk height, node, and boll count. The instance segmentation of the main stalk and nodes was proved to be a challenging task for supervised methods due to the lack of sufficient annotated data and feature similarity in the cotton plant architecture. In order to find a solution to the challenge of data scarcity, two weakly supervised approaches: Weakly Supervised Counting (WS-COUNT) and CountSeg, were demonstrated to carry out the cotton boll counting task. The results showed that weakly supervised counting approaches based on peak response maps such as CountSeg (RMSE = 1.284 ±0.08) yield comparable results as of Mask R-CNN (RMSE = 1.175 ±0.20). In terms of data annotation, the weakly supervised approaches were found to be at least 10 times efficient compared to the supervised approach for the boll counting task. In the future, the weakly supervised approach allows us to improve the current supervised frameworks in the absence of quality mask annotations by leveraging the density maps obtained from weak supervision and can be extended to various cotton organs or other crops.